Improvement of Land Cover / Land Use Classification by
Combination of Optical and Microwave Remote Sensing Data
Nguyen Dinh Duong
Department of Environmental Information Study and Analysis
Institute of Geography
18 Hoang Quoc Viet Rd., Hanoi, Vietnam
Fax: 84-4-8361192, Phone: 84-4-7562417, Email: duong.nd@hn.vnn.vn
Abstract: Optical and microwave remote sensing data have been
widely used in land cover and land use classification. Thanks to
the spectral absorption characteristics of ground object in visible
and near infrared region, optical data enables to extract different
land cover types according to their material composition like
water body, vegetation cover or bare land. On the other hand,
microwave sensor receives backscatter radiance which contains
information on surface roughness, object density and their 3-D
structure that are very important complementary information to
interpret land use and land cover. Separate use of these data have
brought many successful results in practice. However, the
accuracy of the land use / land cover established by this
methodology still has some problems. One of the way to improve
accuracy of the land use / land cover classification is just
combination of both optical and microwave data in analysis. In
this paper for the research, the author used LANDSAT TM scene
127/45 acquired on October 21, 1992, JERS-1 SAR scene
119/265 acquired on October 27, 1992 and aerial photographs
taken on October 21, 1992. The study area has been selected in
Hanoi City and surrounding area, Vietnam. This is a flat
agricultural area with various land use types as water rice,
secondary crops like maize, cassava, vegetables cultivation as
cucumber, tomato etc. mixed with human settlement and some
manufacture facilities as brick and ceramic factories. The use of
only optical or microwave data could result in misclassification
among some land use features as settlement and vegetables
cultivation using frame stages. By combination of multitemporal
JERS-1 SAR and TM data these errors have been eliminated so
that accuracy of the final land use / land cover map has been
improved. The paper describes a methodology for data
combination and presents results achieved by the proposed
approach.
Keywords: Land cover, Classification, Optical and Microwave
data, Remote sensing, Combination.
1. Introduction
Land cover mapping has been one of activities of
utilization of remote sensing data in earth sciences for a
long time. Both optical and microwave remote sensing
data were intensively used for this purpose. Thanks to the
spectral absorption characteristics of ground object in
visible and near infrared region, optical data enables to
extract different land cover types according to their
material composition like water body, vegetation cover or
bare land. On the other hand, microwave sensor receives
backscatter radiance which contains information on surface
roughness, object density and their 3-D structure that are
very important complementary information to interpret
land use and land cover. Separate use of these data have
brought many successful results in practice. However, the
accuracy of the land use / land cover established by this
methodology still has some problems. One of the way to
improve accuracy of the land use / land cover classification
is just combination of both optical and microwave data in
analysis. The ways of combination of these two data types
are several. According to [2] one can fuse optical and
microwave data by pixel, feature and decision based
approach. In pixel based fusion, the data are merged on a
pixel-by-pixel basis. Feature based approach always merge
the different data sources at the intermediate level. Each
image from different sources is segmented and the
segmented images are fused together. Decision based
fusion, the outputs of each of the single source
interpretation are combined to create a new interpretation.
Another concept is usage SAR data as complementary
information source for interpretation of optical imagery
[1]. In this study, the author presents a decision based
combination of Landsat TM and JERS-1 SAR data for land
cover mapping of Hanoi City and surrounding area. A
comparison study of using only optical, single data SAR
and multi-temporal SAR data for land cover mapping have
been carried out. Finally a method for combination of
optical and microwave data has been proposed. Land cover
map which was compiled by the proposed method seems
to be the best because it separates well many land cover
categories that can not be classified if only optical or
microwave data were used.
2. Study Area
The study area is Hanoi City and surrounding area that
is located in north of Vietnam. Land cover of the study
area contains mainly categories as: urban, built-up, rural
human settlement, perennial and short term agricultural
cultivation, rice crop, water bodies etc. The terrain is flat
so there is no need of ortho-rectification of SAR data. Both
optical and SAR can be resample to the same spatial
resolution and overlaid each on the other. On figure 1 is
outline map of the study area.
 Fig. 1. Location of study area in North of Vietnam Fig. 2. Landsat TM false color composite of study area

3. Materials and Methods

The data used for this research consists of Landsat TM

scene 127/45 acquired on October 21, 1992, March 8, 1993
and August 18, 1993. Aerial photos taken on October 21,
1992 were used as ground truth information. Before
analysis, all the image data were geometrically corrected
using topographical map 1:100000. The accuracy of
correction is within one pixel limitation. The JERS-1 SAR
data were resampled to 30m resolution as the same of
Landsat TM data. Fig.3. Land cover of study area established by TM data
and ML classification
3.1 Classification of Landsat TM data The single date JERS-1 SAR data provides information
on land cover through backscatter radiance. The
The Landsat TM data was classified using Maximum understanding of this image requires auxiliary information
likelihood method. Training samples were selected for the and local knowledge of land cover. Solely usage of SAR
following classes: data for land cover still faces difficulties. A simple analysis
has been applied for single date SAR data. The analysis
Sand Sand, sandy beach located along Red was confined to usage of backscatter values for
river classification of land cover objects. On figure 4 is shown
Veget1 Agricultural cultivation and secondary level sliced SAR image of Oct. 21, 1992. The land cover
crop as maize, berry, sugarcane, was separated one from the other mainly by their surface
Veget2 Vegetation with canopy structure structure and not by material composition as in case of
Built Built up area, urban and construction optical sensor. Five categories were classified and they are:
water water body as lake, pond, reservoir To 1 Very flat object. In the study area, it represents
and river mainly water body
wetld Wetland as shallow ponds, flooded To 2 Paddy field, flat object
rice field, irrigated paddy To 3 Low density or light structural object
Agrld1 Harvested paddy field To 4 Medium density or moderate structural object
Agrld2 Paddy field with young secondary To 5 high density or complicated structural object
crop
After selecting training samples, MLC (Maximum
likelihood classification) has been carried out. Figure 2
shows false color composite of study area and figure 3
displays MLC result.
The classified image shows, in general, well distributed
land cover categories. However, there are some
misclassifications in classes built up and agricultural land.
It was also impossible to separate rural human settlement
and perennials that were merged into class Veget2.

3.2 Analysis of single date JERS-1 SAR data Fig. 4. Level sliced JERS-1 SAR image
3.3 Change extraction using multitemporal JERS-1
SAR data
The multitemporal JERS-1 SAR dataset consists of three
dates: Oct. 21, 1992, Mar. 8, 1993 and Aug. 18, 1993. The
dates were chosen so that seasonal variability in land cover
(particularly agricultural cultivation) is maximal. Color
composite of three SAR data is shown on figure 5. Color
assignment is Red= Aug. 18, 1993, Green= Mar. 8, 1993
and Blue= Oct. 21, 1992. Land cover objects that changes
seasonally are shown in different color shades. Stable
objects are displayed in gray tone. It is very easy to extract
urban, rural human settlement and perennials from
agricultural cultivation by visual interpretation. This
interpretation can be carried out digitally by conversion the
color composite to IHS image and extracting uncolored
objects from colored ones. The figure 5 shows color
composite of multitemporal SAR dataset and stable objects
extracted from that are shown on figure 6.
Fig. 5. Color composite of multitemporal SAR dataset
Fig. 6. Stable objects extracted from multitemporal SAR
data
3.4 Land cover classification by combination of SAR
and optical data
Obviously, solely application of optical data as stated in
previous paragraph enables to establish good land cover
map, however, there are still some misclassification and
the result needs to be refined for practical use. Combining
analysis results of both optical and JERS-1 SAR we could
obtain the best result. The figure 7 shows final land cover
map that was established using both optical and SAR data.
Fig. 7. Land cover map established by combination of
optical and microwave data
Beside the classes that were segmented using only optical
data, new classes coming from SAR data were added and
they are:
rhsetl rural human settlement
hdenst high density construction area
mdenst medium density construction area
ldenst low density construction area
agrld3 non flat paddy field ( furrows, beds)
veget4 broad leave agricultural cultivation as edible
canna
4. Conclusion
The decision based combination of optical and
mutitemporal SAR data is quite simple but effective
because it enables to extract rural human settlement,
different agricultural cultivation types, urban and built up
area with different construction density that are impossible
to be classified if only optical or SAR data was used. The
approach presented in this study does not requite any
special software therefore it can be applied even in very
modestly equipped remote sensing laboratory.
Acknowledgement
The author would like to acknowledge the Basic
Research Program of Vietnam for funding this research.
Thanks also belong to staff of Department of
Environmental Information Study and Analysis for
collaboration in field work and data analysis.
References
[1] Z. A. Hasan, K.M.N. Ku Ramil, I. Selamat and K.F. Loh.
Complementary Nature of SAR and Optical Data for Land
Cover/Use Mapping in the State of Johore, Malaysia.
Proceedings of the ACRS 1997.
[2] Jirathana Worawattanamateekul, Canisius X.J.F., Lal
Samarakoon. Multisource data fusion-fusing optical and sar
data for irrigated rice areas identification. Proceedings of the
ACRS 2000.